Postoperative pancreatic fistula affects recurrence-free survival of pancreatic cancer patients

Purpose Postoperative pancreatic fistula (POPF) with reported incidence rates up to 45% contributes substantially to overall morbidity. In this study, we conducted a retrospective evaluation of POPF along with its potential perioperative clinical risk factors and its effect on tumor recurrence. Methods Clinical data on patients who had received pancreatoduodenectomy (PD), distal pancreatectomy (DP), or duodenum-preserving pancreatic head resection (DPPHR) were prospectively collected between 2007 and 2016. A Picrosirius red staining score was developed to enable morphological classification of the resection margin of the pancreatic stump. The primary end point was the development of major complications. The secondary end points were overall and recurrence-free survival. Results 340 patients underwent pancreatic resection including 222 (65.3%) PD, 87 (25.6%) DP, and 31 (9.1%) DPPHR. Postoperative major complications were observed in 74 patients (21.8%). In multivariable logistic regression analysis, POPF correlated with body mass index (BMI) (p = 0.025), prolonged stay in hospital (p<0.001), high Picrosirius red staining score (p = 0.049), and elevated postoperative levels of amylase or lipase in drain fluid (p≤0.001). Multivariable Cox regression analysis identified UICC stage (p<0.001), tumor differentiation (p<0.001), depth of invasion (p = 0.001), nodal invasion (p = 0.001), and the incidence of POPF grades B and C (p = 0.006) as independent prognostic markers of recurrence-free survival. Conclusion Besides the known clinicopathological risk factors BMI and amylase in the drain fluid, the incidence of POPF correlates with high Picrosirius red staining score in the resection margins of the pancreatic stumps of curatively resected pancreatic ductal adenocarcinoma (PDAC). Furthermore, clinically relevant POPF seems to be a prognostic factor for tumor recurrence in PDAC.

Introduction Pancreatoduodenectomy (PD) is the preferred therapeutic method for treating benign and malignant diseases of the pancreatic head and its periampullary region. PD remains a complex and highly invasive visceral resection with mortality rates of about 5% and morbidity rates of up to 60% in experienced high-volume centres [1][2][3]. In line with the demographic changes resulting in an increasing number of multimorbid elderly patients in Europe and the USA, preoperative risk factor assessment and patient stratification have become increasingly popular. Postoperative pancreatic fistula (POPF) with reported incidence rates between 3% and 45% substantially contributes to overall morbidity with increased hospital stay, costs and reintervention rates and in case of abscess formation, sepsis, and hemorrhage to mortality as well [4,5]. Moreover, clinically relevant POPF with prolonged hospital stay may delay adjuvant treatment and affect oncologic outcomes in malignant pancreatic diseases [6].
The crucial importance of standardized reporting of procedures and their complications with uniform definition and classification of POPF and widely supported recommendations for its diagnosis and treatment led to a first consensus statement of the International Study Group of Pancreatic Surgery (ISGPS) in 2005 [7]. In 2017, clinically relevant POPF was redefined as drainage fluid of any measurable volume with amylase level more than three times that of physiological serum amylase activity, associated with a clinically relevant condition related directly to the POPF, which can originate from pancreaticoenteric or pancreaticogastric anastomosis after head resection or drainage procedures and of pancreatic remnant after distal pancreatectomy or enucleation [8].
The efficacy of anastomotic techniques like pancreaticojejunostomy versus pancreaticogastrostomy, invagination versus duct-to-mucosa, internal versus external pancreatic duct stenting, fibrin glue versus other topical haemostatic occlusive agents to seal the pancreatic anastomosis, and the use of various somatostatin analogues to decrease pancreatic enzyme secretion were investigated in various trials [9][10][11][12][13][14][15][16]. So far, there is no consensus on the optimal management of POPF and no standardized intra-and perioperative treatment [17]. Therefore, the attention has been focused on the assessment of POPF risk factors. Many studies addressed the role of single factors or comprehensive risk scores on multiple risk factors, such as age, gender, body mass index (BMI), pathologic diagnosis, operative time, blood loss, diameter of the main pancreatic duct, texture of pancreatic parenchyma, American Society of Anaesthesiologists (ASA) score, heart rate, systolic blood pressure, haemoglobin, and albumin levels [18]. POPF risk stratification customized for individual patients may help in increasing the number of patients eligible for pancreatic surgery by preoperative nutritional support, optimizing cardiovascular medication, starting exercise therapy, and postoperative intensive monitoring or even to select high-risk patients who might be excluded from surgical resection. However, current guidelines for pancreatic surgery do not recommend any of the proposed clinical risk prediction scores for morbidity and mortality such as ASA and POSSUM (Physiological and Operative Severity Score for the enUmeration of Mortality and Morbidity) nor POPF risk scores such as PREPARE (Preoperative Pancreatic Resection) score by Uzunoglu et al., FRS (Fistula Risk Score) by Callery et al., and a-FRS (Alternative Fistula Risk Score for Pancreatoduodenectomy) by Mungroop et al. [19][20][21][22][23].
In this study, we conducted a retrospective evaluation of POPF along with its potential perioperative clinical and histopathological risk factors and its impact on tumor recurrence. postoperative day (POD) 3. Postoperative incidence and treatment of complications, such as pneumonia, heart attack, pulmonary embolism, thrombosis, urinary tract infection, wound infection, delayed gastric emptying, POPF, bile and chyle leakage, abdominal haemorrhage, and abscess were documented. Further data were obtained concerning the length of postoperative hospital stay particularly in the intensive care unit, reoperation, hospital mortality, the histopathological diagnosis, adjuvant chemotherapy regimens, tumor relapse, and tumor related deaths. In total, 72 variables were evaluated in this study and are presented in Table 1.

Definitions of postoperative complications
Major complications were defined according to the Clavien-Dindo classification as grades IIIb to V, validated for pancreatic resections [25,26]. Grade III includes complications requiring surgical, endoscopic or radiological intervention, followed by grade IV life-threatening complications requiring intensive care unit (ICU) management, and grade V as the complicationrelated death of the patient. POPF was defined according to the reclassification of the ISGPS in 2017 [8]. Depending on the clinical course three grades of severity were defined: 1. Biochemical leak (BL), former POPF grade A, with increased amylase activity in the fluid of a drain from POD 3 but without clinical implications. 2. POPF grade B that requires a change in the expected postoperative management with prolonged drainage, therapeutic agents and lessinvasive treatment including percutaneous, endoscopic, or angiographic interventional procedures. 3. A shift to POPF grade C is characterized by organ failure, clinical instability, and high mortality, requiring reoperation and stay in an ICU.
Postoperative bile leakage was defined according to the definitions of the International Study Group of Liver Surgery (ISGLS) as discharge of fluid of an intra-abdominal drain with a bilirubin concentration at last three times greater than the serum bilirubin persisting on POD 3 [27]. Chyle leak after pancreatic resection was defined as the output of milky-colored fluid of a drain from POD 3 with a triglyceride content �110mg/dl [28]. In accordance with the definitions of the ISGPS, post-pancreatectomy hemorrhage was defined according to the time of onset, location, and clinical severity [29]. Delayed gastric emptying was diagnosed if the patient was unable to tolerate a solid diet on POD 7 or if reinsertion of the nasogastric tube was required between POD 4 and 7 according to the definitions of the ISGPS [30]. Mortality was defined as death occurring during the hospital stay.

Histochemical staining by picrosirius red
The resection margins of pancreatic stumps of all cases treated by pancreatic head resection were submitted for pathological diagnosis by classic H&E staining. In addition, 6μm thick formalin-fixed paraffin embedded (FFPE) tissue sections of the resection margins were deparaffinized through xylene and a graded alcohol series. Tissue samples were stained for 30 min in Picrosirius red solution consisting of 0.1% solution of Sirius Red F3B (Sigma-Aldrich, St Louis, MO, USA) in saturated aqueous picric acid [31]. The stained sections were then washed for 2 min in 0.01 N HCI, dehydrated, cleared and mounted in synthetic resin. Negative controls without Sirius Red and positive controls (liver fibrosis and cirrhosis) were included in all experiments using the same experimental conditions. Picrosirius red planar staining was evaluated separately by two independent investigators in a blinded manner using light microscopy (Eclipse E1000M and NIS-Elements D3.1 imaging software, Nikon, Tokyo, Japan) with a modified score range from 1 to 3 according to Ridolfi et al. and Gaujoux et al. [32,33] Three grades of fibrosis were defined: 1. Normal pancreatic parenchyma, consisting in lobes separated by connective tissue organized in fine septa. 2. focal perilobular or periacinar fibrosis (Fig 1A), and 3. complete replacement of the parenchyma by fibrosis with rare residual areas of acinar glandular tissue. Similarly, the grade of inflammation by lymphocytes and of fatty infiltration by adipocytes was classified in three grades as absent, focal, and generalized (Fig 1B). Based on this, a global scoring system for morphological classification of the resection margin of the pancreatic stump was developed. Resulting from the sum of the values for each of the three morphological features of fibrosis, inflammatory and fatty infiltration, three groups A-C were defined: A: resection margin of the pancreatic stump with no or marginal alterations of inflammation and steatosis but generalized fibrosis (3)(4), B: resection margin of the pancreatic stump with moderate alterations (5-6), and C: resection margin of the pancreatic stump with severe alterations of inflammation and steatosis but absent or marginal fibrosis (7-9).

Statistical analysis
Statistical analyses were performed by SPSS Statistics Version 22 (IBM Corp. Armonk, NY) for Windows. Categorical parameters are reported as absolute and relative frequencies and  continuous parameters as median [minimum-maximum]. All continuous variables were considered in a dichotomized form in further analyses. The association between clinicopathological parameters and the incidence of (no, non-severe, major) complications, and (no, BL, B, C) POPF was assessed by Fisher's exact test. Uni-and multivariable logistic regression analyses were performed to identify independent risk factors for grade B/C POPF and major postoperative complications. All parameters with p�0.05 were included in the multivariable analyses.
Results are reported as odds ratios (OR), corresponding 95% confidence intervals (CI) and p-values.
Overall survival and recurrence free survival were measured from the date of surgery to the time of the last follow-up or cancer-related death or tumor relapse, respectively, considering patients who were still alive or without evidence of tumor relapse at the end of the study as censored. The Kaplan-Meier method was used to generate overall and recurrence-free survival curves of curatively treated PDAC patients with and without major postoperative complications and grade B/C POPF. Survival curves were compared by the log-rank test. A Cox proportional-hazards regression model was used to assess the impact of potential risk factors on overall and recurrence-free survival. All parameters with p�0.05 were included in the multivariable models. A forward stepwise variable selection procedure based on the likelihood ratio test was applied. Results are reported as hazard ratios (HR), corresponding 95% CIs and p-values. All inferential statistics were intended to be exploratory. P�0.05 was considered statistically noticeable.
Pancreatic anastomosis after PD was performed by duct-to-mucosa and end-to-side pancreaticojejunostomy in all patients. A surgical drain was placed in all patients. PD with       The mean length of hospital and intensive care unit stay was 20 and 6 days, respectively. The 30 days mortality rate was 5% (n = 17) in total, 6.3% (n = 14) in the PD, 3.4% (n = 3) in the DP, and 0% in the DPPHR group.
Among patients that developed a POPF grade B (n = 30; 8.8%), a total of 28 (93.3%) were treated by computed tomography guided percutaneous drainage for infected intraabdominal fluid collections and 3 angiographic procedures with successful coil embolization of a ruptured gastroduodenal artery pseudoaneurysm were performed. Surgical reoperations were required in 31 patients with POPF grade C (9.1%). In detail, 12 cases (38.7%) with recreation of the pacreaticojejunostomy, 4 cases (12.9%) with recreation of the pancreaticojejunostomy as well as the biliodigestive anastomosis, 7 cases (22.6%) with recreation of the pancreaticojejunostomy as well as the biliodigestive anastomosis followed by relaparotomy with resection of the pancreatic stump, 3 cases (9.7%) with primary resection of the pancreatic stump, 4 cases (12.9%) with intraabdominal haematoma removal and resection of the bleeding pseudoaneurysm of the gastroduodenal artery stump, and 1 case (0.9%) with relaparotomy and removal of abdominal abscess formation were performed as postoperative complication management.

Picrosirius red staining score
Histochemical staining by Picrosirius Red and its grading of fibrosis, inflammation, and steatosis were performed in pancreatic tissue sections of the resection margins of 178 consecutive study patients with PD since 2009. Table 8 shows the distribution of the Picrosirius red global score and each of the three morphological features the score is based on in stained tissue samples by POPF grading. Low-grade fibrosis and high-grade inflammation or steatosis of the pancreatic remnant were associated with higher POPF grades (each p<0.01). A correlation between the Picrosirius red staining score and the incidence of severe postoperative complications and clinically relevant POPF was also found in the univariable logistic regression analyses (p�0.001) (Tables 3 and 4). In multivariable logistic regression analyses, the Picrosirius red staining score correlated only with POPF (p = 0.049). Cox regression analysis for overall and  (Tables 6 and 7).

Discussion
In the last decades, advances in surgical techniques and in perioperative care, in addition to careful patient selection, have improved the outcomes in pancreatic surgery with considerable decrease of mortality rates below 5%. However, morbidity remains close to 50% even in highvolume centres, where clinically relevant POPF in 12 to 20% of patients continues to be the major cause of subsequent severe complications [2,3,34]. Numerous studies have identified a number of risk factors associated with an increased incidence of POPF including age, gender, BMI, pancreatic duct diameter, pancreas texture, operative time, and anastomosis technique [3,19,20,22,33]. Logistic regression analysis of clinicopathological parameters in our study cohort could confirm increased risk of major complications post pancreatic surgery in patients with advanced age, increased BMI and ASA, coexisting cardiopulmonary diseases, absence of preoperative chronic pancreatitis, intraoperative blood transfusion, prolonged operative and anaesthesia time, PD procedure, soft pancreatic tissue, and absence of internal or external pancreatic duct stenting. Furthermore, preoperative anemia and elevated serum levels of bilirubin and CRP were associated with postoperative complications. The multivariate correlation of Bold values indicate significance (p� 0.05). 1 HR, hazard ratio postoperative major complications and elevated preoperative serum levels of CRP �8 mg/l reflects the systemic inflammatory response of the study patients with predominantly malignant and chronic inflammatory pancreatic diseases. However, the results suggest that even a subtle rise in preoperative CRP may stratify patients into a higher risk group for postoperative major complications. A recent study by Oehme et al. identified preoperative anemia as a risk factor for postoperative complications greater than grade 2 according to Clavien-Dindo and as an independent prognostic factor for shorter overall survival in patients undergoing surgical procedures for pancreatic malignancies [35]. A preoperative serum bilirubin level of >3 mg/dl was the most significant risk factor for clinically relevant POPF in a retrospective analysis by Rungsakulkij et al. [36] Univariable regression analysis revealed that the rise in the serum levels of amylase and lipase in the early postoperative period was associated with postoperative major complications suggesting manifestation of intraoperatively induced self-resolving pancreatitis [37]. In line with a multitude of reports, elevated amylase activity in the drainage fluid on POD 3 correlated with postoperative major complications, particularly with POPF [2,5,7,8,18,38,39]. The POPF definition by the ISGPS using amylase drain concentration is widely used. However, in accordance with our findings Tzedakis et al. could show that lipase is as effective as amylase drain concentration to define POPF [40]. Consequently, ICU and hospital stay were prolonged in patients with postoperative complications. In particular, the incidence of clinically relevant POPF correlated with a high score of histological Picrosirius red staining in multivariable analysis. The used Picrosirius red F3BA is a strong, linear anionic dye comprising six sulfonate groups that can associate along cationic collagen fibres. Picrosirius red staining developed by Junqueira et al. [31] provides a simple, specific, and sensitive method for localizing fibrillar collagen in tissue sections. In addition to the surgeon's judgment about pancreatic texture by palpation, this histological tool allows a more objective assessment of the pancreatic remnant before reconstruction. There is broad agreement that fibrosis and lowgrade fatty or inflammatory cell infiltration of the pancreatic parenchyma are POPF protective factors. It could be demonstrated that increased pancreatic fibrosis e.g. due to chronic pancreatitis not only enables a more secure anastomosis with solid fixing sutures but was also associated with decreased exocrine activity with reduced pancreatic juice output [32,33,[41][42][43]. Therefore, intraoperative knowledge about the pancreatic texture in addition to the dignity of the resection margin of the pancreatic stump could help surgeons to adjust their surgical procedure. Besides the choice of the anastomotic technique, the use of intraoperative internal or external pancreatic duct stenting or even total pancreatectomy to prevent devastating POPF could be the consequence. We believe that further prospective studies evaluating the impact of intraoperative frozen section histology on anastomotic technique and outcome could help pancreatic surgeons prevent clinically relevant POPF. In addition to patient characteristics and blood tests, the preoperative risk assessment of non-invasive cross-sectional imaging related to POPF has been the focus of several research groups [44,45]. A systematic review and meta-analysis of POPF prediction using preoperative computed tomography (CT) scan by Yue et al. revealed a significant increase in the incidence of clinically relevant POPF in patients with visceral obesity and sarcopenic obesity [46]. Furthermore, a narrow pancreatic duct assessed by preoperative CT images significantly related to POPF [47]. Multiparametric magnetic resonance imaging of the pancreas by Yoon et al. could enable accurate quantification of pancreatic fibrosis and steatosis, which have been shown to be associated with POPF [48]. The correlation of preoperative cross-sectional imaging with the surgeon's judgment about pancreatic texture by palpation and intraoperative histological findings may allow a more objective assessment of the pancreatic remnant to prevent POPF and should be evaluated in future studies.
Although high grade POPF is universally regarded as a major source of early postoperative morbidity and mortality, its role in oncological outcome remains uncertain. Indeed, only a few studies have investigated the impact of clinically relevant POPF on PDAC specific survival and recurrence with contradictory results [6,[49][50][51]. However, recent reports have demonstrated that early initiation of adjuvant chemotherapy was an important prognostic factor in patients with PDAC as severe POPF significantly prolonged initiation of adjuvant chemotherapy after primary surgery [52,53]. In curative resected PDAC patients of our study, Cox regression analyses indicated a survival benefit for low UICC and NYHA stages, low grading, low tumor invasion, absence of nodal invasion, low preoperative serum level of CA.19-9, and low incidence of postoperative major complications according to Clavien-Dindo. Advanced UICC stages and tumor differentiation, tumor invasion, nodal invasion, high preoperative serum level of CA.19-9, and the incidence of major complications, in particular clinically relevant POPF were identified as potentially favorable factors for tumor recurrence. Moreover, high grade POPF correlated with recurrence-free survival in multivariable analysis. Our results are in line with the negative influence of anastomotic leakage on survival outcome with high incidence of local recurrence in other gastrointestinal carcinoma entities [54]. It is supposed that anastomotic leakage leads to inflammation with the release of pro-inflammatory cytokines that alter host defense and promote growth of residual cancer cells [55]. In PDAC, Nagai et al. were able to demonstrate that clinically relevant POPF was an independent prognostic factor for peritoneal tumor recurrence [51]. Interestingly, even 28% of R0 curative resected patients with invasive PDAC revealed postoperatively higher cytology-positive rates in the drained fluid from the pancreatic bed with subsequent development of local recurrence [56].
Taken together, in addition to known risk factors the Picrosirius red staining comes along with a high diagnostic potential in the risk management of POPF that correlates with tumor recurrence in this study cohort. Further multicenter studies with a larger number of PDAC patients are required to reevaluate the value of intraoperatively extended histological diagnostic as well as the impact of high grade POPF as an independent survival predictor.

Conclusion
This study indicates the high potential of pre-surgical risk stratification of known clinical risk factors and the intraoperative histopathological diagnostic of fibrosis, fatty and inflammatory cell infiltration in the resection margins of pancreatic stumps of curatively resected PDAC patients to prevent devastating POPF. Its prevention is urgently needed as clinically relevant POPF seems to be a prognostic factor of tumor recurrence in PDAC.